Building an API Bridge

Using Techcyte's classification webhook system, third-party developers can run their image classification code on their own infrastructure and report results back to Techcyte. When a user creates an image classification task, the webhook is notified with the appropriate variables listed below.

Example Code Features:

Modular Code: Implement your image processing logic in webserver.py’s process_image() function.
Image Handling: Support DICOM/SVS/TIFF via pydicom, openslide-python
Visualization: On the Techcyte app, four box objects are drawn on the image for result verification.

Webhook variables

Supplied in Production:

company_id: Company identifier, useful for billing
scans: A mapping of scan identifiers to presigned image download url
task_id: Task identifier
case_id: The assigned case id (unused for most calls)
model_id: A user supplied variable used to customize webhook calls
jwt_token: A task specific jwt token used for authorizing requests to techcyte (not required for local testing)

Getting Started

1. Clone the Repository:

git clone https://github.com/Techcyte/devkit.git
cd devkit/src/api-bridge

2. Run the example:

You'll need to generate an API key and specify when running the webserver. See creating an API key

Start the webserver

# Install requirements with
# pip install -r requirements.txt
# openslide-bin may be required depending on your system
# pip install openslide-bin
python webserver.py --port 3000 --api-key-id e-_tfr-redacted-Vhjt --api-key-secret FNI-redacted-LvH

Mock a techcyte webhook request. See creating a debug request for test data variables.

 curl -X POST --url "http://localhost:3000/webhook" \  
 --header "Content-Type: application/json" \
 --data '{ \
    "company_id": "2380941", \
    "scans": { \
      "8823846": "https://techcyteci-preprod.s3.us-west-2.amazonaws.com/redacted" \
    }, \
    "case_id": "aHVtYW5DYXNlOjI0MjEzNjY=", \
    "task_id": "dGFzazoxODU4MzU=", \
    "model_name": "" \
  }'

3. Customize Your Code:

Edit webserver.py, replacing the process_image() function with your classifier logic.
Ensure the output matches the required schema (see below).

It is possible to run this example with ngrok (https://ngrok.com) and process Techcyte images locally for testing (ngrok http 3000). But a more robust solution should be implemented for production environments.

4. Deploying

See creating a debug request for setting up your endpoint in the Techcyte app. The only difference is you will fill in the webhook url with your publicly available url.

Posting Results

You will need to modify the caseResults and/or scanResults values to see your results on the Techcyte app. See posting external model results.

Implementation tips

Running a GPU classifier non-stop can be expensive. Consider using a system like AWS Lambda (triggered by an API bridge event) and AWS Batch to process requests efficiently.
Images may be in SVS, TIFF, or DICOM ZIP format, handle as needed.